search

zhangyp15 / OccFormer

[ICCV 2023] OccFormer: Dual-path Transformer for Vision-based 3D Semantic Occupancy Prediction
https://arxiv.org/abs/2304.05316
Apache License 2.0
324 stars 22 forks source link

Have you ever try origin bev_encoder？ #5

Closed LMerCy closed 1 year ago

LMerCy commented 1 year ago

Have you ever try origin bev_encoder of bevdepth, how about it's performance?

zhangyp15 commented 1 year ago

Yes, the bev-encoder from bevdepth performs slightly worse than the transformer encoder. You can refer to the 3D ResNet-16 entry in the ablation study on encoders (Table 4).

LMerCy commented 1 year ago

Yes, the bev-encoder from bevdepth performs slightly worse than the transformer encoder. You can refer to the 3D ResNet-16 entry in the ablation study on encoders (Table 4).

Thanks for your work! Could you provide a config of this?

zhangyp15 commented 1 year ago

Sure.

You can check occformer_3dconv.zip, which includes the config, the 3d resnet module, and our training log from the previous implementation.

LMerCy commented 1 year ago

occformer_3dconv.zip

Thanks!